PNNL-21283

Prepared for the U.S. Department of Energy under Contract DE-AC05-76RL01830

R. César Izaurralde Ritvik Sahajpal Xuesong Zhang David H. Manowitz April 2012

PNNL-21283

National Geo-Database for Biofuel Simulations and Regional Analysis of Biorefinery Siting Based on Cellulosic Feedstock Grown on Marginal Lands

R César Izaurralde Ritvik Sahajpal

Xuesong Zhang David H. Manowitz

April 2012

Prepared for

the U.S. Department of Energy

under Contract DE-AC05-76RL01830

Pacific Northwest National Laboratory

Richland, Washington 99352

iii

Executive Summary

The goal of this project undertaken by GLBRC (Great Lakes Bioenergy Research Center) Area 4

(Sustainability) modelers is to develop a national capability to model feedstock supply, ethanol production,

and biogeochemical impacts of cellulosic biofuels. The results of this project contribute to sustainability goals

of the GLBRC; i.e. to contribute to developing a sustainable bioenergy economy: one that is profitable to

farmers and refiners, acceptable to society, and environmentally sound. A sustainable bioenergy economy

will also contribute, in a fundamental way, to meeting national objectives on energy security and climate

mitigation.

The specific objectives of this study are to: (1) develop a spatially explicit national geodatabase for

conducting biofuel simulation studies and (2) locate possible sites for the establishment of cellulosic ethanol

biorefineries.

To address the first objective, we developed SENGBEM (Spatially Explicit National Geodatabase for

Biofuel and Environmental Modeling), a 60-m resolution geodatabase of the conterminous USA containing

data on: (1) climate, (2) soils, (3) topography, (4) hydrography, (5) land cover / land use (LCLU), and (6)

ancillary data (e.g., road networks, federal and state lands, national and state parks, etc.). A unique feature of

SENGBEM is its 2008-2010 crop rotation data, a crucially important component for simulating productivity

and biogeochemical cycles as well as land-use changes associated with biofuel cropping.

ARRA support for this project and to the PNNL Joint Global Change Research Institute enabled us to

create an advanced computing infrastructure to execute millions of simulations, conduct post-processing

calculations, store input and output data, and visualize results. These computing resources included two

components installed at the Research Data Center of the University of Maryland. The first resource was

“deltac”: an 8-core Linux server, dedicated to county-level and state-level simulations and PostgreSQL

database hosting. The second resource was the DOE-JGCRI “Evergreen” cluster, capable of executing

millions of simulations in relatively short periods. ARRA funding also supported a PhD student from UMD

who worked on creating the geodatabases and executing some of the simulations in this study.

Using a physically based classification of marginal lands, we simulated production of cellulosic

feedstocks from perennial mixtures grown on these lands in the US Midwest. Marginal lands in the western

states of the US Midwest appear to have significant potential to supply feedstocks to a cellulosic biofuel

industry. Similar results were obtained with simulations of N-fertilized perennial mixtures. A detailed spatial

analysis allowed for the identification of possible locations for the establishment of 34 cellulosic ethanol

biorefineries with an annual production capacity of 5.6 billion gallons.

In summary, we have reported on the development of a spatially explicit national geodatabase to conduct

biofuel simulation studies and provided simulation results on the potential of perennial cropping systems to

serve as feedstocks for the production of cellulosic ethanol. To accomplish this, we have employed

sophisticated spatial analysis methods in combination with the process-based biogeochemical model EPIC.

The results of this study will be submitted to the USDOE Bioenergy Knowledge Discovery Framework as a

way to contribute to the development of a sustainable bioenergy industry. This work provided the opportunity

to test the hypothesis that marginal lands can serve as sources of cellulosic feedstocks and thus contribute to

avoid potential conflicts between bioenergy and food production systems. This work, we believe, opens the

iv

door for further analysis on the characteristics of cellulosic feedstocks as major contributors to the

development of a sustainable bioenergy economy.

v

Acknowledgments

We gratefully acknowledge the support from the following:

DOE EERE Office of Biomass Program: Alison Goss, Ranyee Chiang

DOE EERE Golden Field Office: Steven Thomas

PNNL: Jonathan Male, Don Stevens, Gregg Andrews, Paul Girgis, Kay Killingstad, Kali Wood

ORNL: Stan Wullschleger, Wilfred Post

USDA – FAS: Rich Iovanna

MSU: Ilya Gelfand, G. Philip Robertson

vi

Acronyms and Abbreviations

ANPP Aboveground Net Primary Productivity

CDL USDA Crop Data Layer

DEM Digital Elevation Model

EERE USDOE Energy Efficiency and Renewable Energy

EISA 2007 Energy Independence and Security Act of the United States of America

EPIC Environmental Policy Integrated Climate model

GIS Geographic Information System

GLBRC Great Lakes Bioenergy Research Center

HCU Hydrologic Catalogue Unit

HDOS Hierarchical Data Organization System

HSMU Homogenous Spatial Modeling Unit

LCA Life Cycle Analysis

LCC Land Capability Class

LCGHGE Lifecycle Greenhouse Gas Emission

NASS USDA National Agricultural Statistical Census

NHD National Hydrography Dataset

NOAA National Oceanographic and Atmospheric Administration

NRC National Research Council of the United States of America

OBP USDOE EERE Office of Biomass Program

RIMA Regionally Intensive Modeling Area

RFS Renewable Fuel Standard, a component of the 2005 Energy Policy Act of the United

States of America

SEIMF Spatially Explicit Integrated Modeling Framework

SENGBEM Spatially Explicit National Geodatabase for Biofuels and Environmental Modeling

SRTM Shuttle Radar Topography Mission

SSURGO USDA Soil Survey Geographic Database

TPI Topographic Position Index

USDA United States Department of Agriculture

USDOE United States Department of Energy

Acronyms and Abbreviations Mg Equivalent to 1 metric ton or 1,000 kg

Gg Equivalent to 1,000 Mg

vii

Contents

Executive Summary .......................................................................................................................... iii

Acknowledgments ............................................................................................................................... v

1.0 Introduction ................................................................................................................................ 1

1.1 Background and Rationale ................................................................................................. 1

1.2 General and Specific Objectives ........................................................................................ 2

2.0 Materials and Methods ............................................................................................................... 2

2.1 Overall Approach ............................................................................................................... 2

2.2 Preparation of a Spatially Explicit National Geodatabase for Biofuel and

Environmental Modeling (SENGBEM) ............................................................................. 3

2.3 Preparing and Implementing Spatially Explicit Simulations with the EPIC Model .......... 5

2.3.1 The EPIC Model ...................................................................................................... 5

2.3.2 Spatially Explicit Modeling Framework ................................................................. 5

2.3.3 Site Validation ......................................................................................................... 6

2.4 Regional Simulations ......................................................................................................... 8

2.4.1 Simulation of Biomass Feedstock across the 10-State US Midwest ....................... 8

2.4.2 Methodology for Siting Biorefineries ..................................................................... 9

3.0 Results and Discussion ............................................................................................................... 9

3.1 Features of the National Geodatabase to Simulate Biofuel Production ............................. 9

3.2 Simulations Results .......................................................................................................... 11

3.2.1 Site Simulations..................................................................................................... 11

3.2.2 Potential Biomass Production from Marginal Lands ............................................ 12

3.2.3 Comparison with Estimates from Billion Ton Study Update ................................ 13

3.2.4 Identifying Location of Biorefineries in the 10-State US Midwest Region .......... 14

3.3 Perennial Biomass Crop Production in the Sandhills of Nebraska .................................. 16

4.0 Conclusions .............................................................................................................................. 18

5.0 References ................................................................................................................................ 20

6.0 Publications and Presentations ................................................................................................. 22

viii

Figures

Figure 1. Structure of the Spatially Explicit Modeling Framework (Zhang et al., 2010).. ................ 2

Figure 2. Illustration of the hierarchical and flexible downscaling and upscaling schemes in

United States lands (Zhang et al., 2010). .................................................................................... 6

Figure 3. Schematic diagram describing the procedures for defining HSMUs (Homogenous

Spatial Modeling Units). ............................................................................................................. 7

Figure 4. National Geodatabase of Land Cover and Use in the Conterminous USA.. ..................... 10

Figure 5. Potential cellulosic ethanol biorefinery locations in the 10-state US Midwest Region.. ... 14

Figure 6. The placement of biorefineries for cellulosic biomass production from marginal lands in

ten North Central states with identification numbers corresponding to Table 3.. ..................... 16

Figure 7. Production capability across the Sandhills for a native prairie mix. The dune ridges and

slopes of the Sandhills have been excluded from analysis.. ..................................................... 17

ix

Tables

Table 1. Average biomass yields (Mg ha-1 yr-1) for 78,184 parcels of marginal land as modeled

by EPIC for three levels of N fertilizer application and two harvest efficiencies.. ................... 13

Table 2. EPIC simulated and Billion-Ton Study Update estimated biomass yields on marginal

lands across the 10-state US Midwest Region.. ........................................................................ 13

Table 3. EPIC modeled biomass and cellulosic ethanol production by state assuming 90% harvest

efficiency and two N-fertilization levels (0 and 68 kg ha-1

).. ................................................... 15

1

1.0 Introduction

1.1 Background and Rationale

Biofuels have been emerging as viable alternatives to fossil fuels. Two major factors have

determined their rise in prominence in the United States: energy security and climate change (NRC,

2011). Energy security concerns about high oil prices and a continued dependence of the U.S. on

foreign oil, led the U.S. Congress to establish the Renewable Fuel Standard (RFS) as part of the 2005

Energy Policy Act. The RFS established the first renewable fuel volume in the U.S. in the form of a

requirement to blend 7.5 billion gallons of renewable fuel with gasoline by 2012. The 2007 Energy

Independence and Security Act expanded the RFS mandate by requiring an increase in the fuel

blending from 9 billion gallons in 2008 to 36 billion gallons by 2022. Of the 36 billion gallons by

2022, about 58% has to originate from advanced biofuels; i.e. a renewable fuel that has a lifecycle

greenhouse gas emission (LCGHGE) that is at least 50% lower than that of the baseline LCGHGE in

2005 (EISA, 2007). Cellulosic ethanol, an advanced biofuel, is expected to cover 44% of the 2022

target.

The second factor relates to climate change and the societal need to reduce anthropogenic

greenhouse-gas (GHG) emissions to the atmosphere. Biofuels are promising because they can

contribute to replace energy use from fossil fuels and reduce GHG emissions to the atmosphere (Kim

and Dale, 2004). Satisfying bioenergy production targets in the U.S. will create a demand for land to

grow biomass crops and some have expressed concern about the ultimate impact of these

transformations on food prices and the environment (Crutzen et al., 2008; Fargione et al., 2008;

Searchinger et al., 2008).

Ensuring the sustainable production of biofuels is a key mission of the Great Lakes Bioenergy

Research Center (GLBRC) created in 2007 by the U.S. DOE Office of Science and led by the

University of Wisconsin and Michigan State University (Slater et al., 2010). The GLBRC is

organized around four major discovery areas: (1) improved plant production, (2) improved

processing, (3) improved catalytical processes, and (4) sustainable biofuels production practices. Six

major research activities describe the sustainability area: novel production practices, plant-microbial

interactions, biogeochemical practices, biodiversity, economic analysis, and integrated modeling

(biophysical, biogeochemical, economic, and LCA).

Most of the integrated modeling research at the GLBRC has been conducted at biorefinery scale,

a multi-county area where a theoretical cellulosic ethanol biorefinery could draw enough plant

biomass to produce 100 million gallons of ethanol per year. These areas are known within the

GLBRC research community as RIMAs (Regionally Intensive Modeling Areas). Zhang et al. (2010)

reported on SEIMF (Spatially Explicit Modeling Framework), a spatially explicit approach designed

to model biomass productivity and environmental impacts of diverse biofuel crops. Egbendewe-

Mondzozo et al. (2011) expanded on the work by Zhang et al. (2010) by developing a spatially

explicit bioeconomic model of biomass supply from alternative cellulosic crops and crop residues as

well as explored policy scenarios for handling various environmental outcomes. Current modeling

work focuses on integrating the SEIMF approach with biogeochemical and biodiversity experiments

together with LCA modeling. Emerging results from field experiments are utilized to test as well as

improve modeling capabilities of current and novel biofuel-production practices.

2

In fall 2008, the USDOE EERE Office of Biomass Program (OBP) provided support to the

GLBRC to enhance field research and modeling activities of sustainable biofuels production

practices. Here we report on efforts to develop a national-scale methodology to simulate biofuel

crops and on regional modeling results of biomass productivity and potential ethanol production

resulting from the collection and processing of this biomass.

1.2 General and Specific Objectives

The objective of this work is to develop predictive land-use change modeling capabilities to

address how biomass can be produced on a sustainable basis. The results of experimental field

research will be combined with theoretical modeling capabilities to create the ability to examine land-

use changes in a broad context with relation to many sustainability criteria such as water use, CO2

emissions, and related factors.

Specific objectives of this report are to describe: 1) a spatially explicit national geodatabase to

conduct biofuel simulation studies and 2) results of a regional analysis of biorefinery siting based on

cellulosic feedstock grown on marginal lands.

2.0 Materials and Methods

2.1 Overall Approach

The overall approach used in this project was

to extend the integrated modeling research

experience gained with the SEIMF approach

(Zhang et al., 2010) at RIMA (multi-county)

scale and extend it for the modeling of larger

(multi-state) regions. As described by Zhang et

al. (2010), the SEIMF contains three modules: 1)

a GIS-based geodatabase, 2) a terrestrial

ecosystem model, and 3) a multi-objective

optimization algorithm (Figure 1). The

geodatabase is used to process input data, define

homogenous spatial modeling units (HSMUs)

from moderate to high spatial-resolution data,

and extract input data needed to execute the

terrestrial ecosystem model EPIC (Williams et

al., 1989; Kiniry et al., 1995), the second module

of the SEIMF. The last module of SEIMF is a

multiobjective optimization algorithm designed

for the evaluation of the production and

environmental tradeoffs of diverse biofuel

production practices.

Figure 1. Structure of the Spatially Explicit Modeling

Framework (Zhang et al., 2010).

3

2.2 Preparation of a Spatially Explicit National Geodatabase for Biofuel and Environmental Modeling (SENGBEM)

The main purpose for developing a high-resolution spatial and temporal geodatabase was to be

able to model, at field scale, the performance of diverse biofuel crops under various management

scenarios. In particular, there was the need to examine the potential of marginal lands for the

placement of perennial biofuel crops. The procedure described by Zhang et al. (2010) demonstrated

the possibility of identifying and modeling the productivity and environmental impacts of biofuel

crops grown on marginal lands.

The SENGBEM geodatabase contains five main types of data: 1) climate, 2) soils, 3) topography,

4) hydrography, and 5) land cover / land use (LCLU). In addition, SENGBEM contains other types of

ancillary data such as road networks, federal and state lands, national and state parks, etc. Following

is a description of the data and datasets used in constructing SENGBEM.

Climate Data. The climate database contains historical daily values of weather variables needed

to drive EPIC and other biophysical models. The weather variables are air temperature (maximum

and minimum, ºC), precipitation (mm), solar radiation (MJ m-2

), wind speed (m s-1

), and relative

humidity (as a fraction). Climate data can be extracted from multiple sources, including Daily Surface

Weather and Climatological Summary (DayMet) (http://www.daymet.org/), the North American

Regional Reanalysis (NARR) (http://www.esrl.noaa.gov/psd/data/gridded/data.narr.html), and the

North American Land Data Assimilation System (NLDAS)

(http://www.emc.ncep.noaa.gov/mmb/nldas/). The datasets differ in terms of spatial resolution and

length of period of record.

The DayMet dataset provides the highest spatial resolution (1 km) and accounts for the influence

of orography on weather. DayMet was developed at the University of Montana, Numerical

Terradynamic Simulation Group (NTSG) to provide the biophysical modeling community the daily

meteorological and climatological data necessary to run simulation models. The NARR and NLDAS

datasets are both from NOAA and the main difference between them is the spatial resolution: 32 km

for NARR and 12 km for NLDAS. The main advantage of the NOAA datasets over DayMet is that

they are available from 1979 to the present, with approximately a 6-month lag. While DayMet data

are also available from about the same start period (1980), the delay for more recent data is much

longer.

Topography and Hydrography Data. Topographic data are essential to delineate watersheds and

determine slope characteristics (elevation, gradient, length, and aspect) needed for hydrological and

erosion modeling. Topographic data were derived from the Shuttle Radar Topography Mission

(SRTM), which produced the highest resolution digital elevation model (DEM) of the Earth at a

resolution of 30 m (Farr et al., 2007). In addition, the National Hydrography Dataset (NHD-plus)

(Simley and Carswell, 2009) was used to extract hydrologic catalogue unit (HCU) boundary

information for modeling soil sediment yield, surface runoff, subsurface flow, and nutrient transport

with runoff.

Soils. There are two major soil databases available in the U.S. for environmental modeling: State

Soil Geographic (STATSGO) and Soil Survey Geographic (SSURGO) both developed by USDA-

NRCS. The STATSGO database was designed mainly for regional, multistate, river basin, state and

4

multicounty resource planning, management, and monitoring. Instead, the SSURGO database was

designed primarily for farm and ranch, landowner/user, township, county, or parish natural resource

planning and management. The main difference between the two is the level of detail contained in the

map units; STATSGO map units can have up to 21 different component soils while SSURGO map

units usually contain one component and can contain up to a maximum of three components. The

SSURGO database, now available for almost the entire conterminous U.S., was selected for building

SENGBEM. Data from SSURGO (http://datagateway.nrcs.usda.gov) were extracted to map soils at

scales of ~1:24,000 (~30 m). Data retrieval also included a suite of physical and chemical soil

properties needed for biophysical and biogeochemical modeling: albedo, layer depth (m), bulk

density (Mg m-3

); mineral fractions (sand, silt, clay, coarse fragments), organic C (%), total N (mg kg-

1); and pH. The SSURGO database provides an interpretive classification variable [i.e. land capability

(LC)] that is represented by land classes based on use limitation (e.g., soil depth, erosion risk, slope,

etc.) (Klingebiel and Montgomery, 1961). There are eight LC classes: I–VIII. Class I is (prime) land

without any limitations for use. Class VIII is land that cannot be used for anything but wildlife.

Classes I–IV can support cropland agriculture, whereas classes V–VIII contain non-arable land. LC

can be used for two purposes: (1) identifying marginal land (e.g. low yield cropland) for biofuel

production and (2) qualitatively validating the results of biophysical and biogeochemical modeling.

Land capability classes are subdivided into land capability subclasses according to the kind of

limitation (susceptibility to erosion, excess water, shallowness of the rooting zone, climate hazard).

All map unit components, including miscellaneous areas, are assigned a capability class and subclass.

Land Cover and Land Use. Crop rotation information (e.g. corn-soybean, continuous corn) is

important for driving cropland ecosystem simulations via a spatially explicit terrestrial ecosystem

model. Land cover / land use data to build SENGBEM were derived from the USDA Cropland Data

Layer (CDL) remote sensing product available at national scale at a resolution of 56m x 56m since

2008 (http://datagateway.nrcs.usda.gov) for the coterminous U.S. The CDL land cover was developed

by the USDA to utilize seasonal satellite imagery to monitor how the acreage of major crop types

varies from one year to the next. To generate a high-resolution crop rotation map of the U.S, multiple

years (2008, 2009, and 2010) of CDL images were overlaid to derive historical crop rotations. A

novel time series analysis based algorithm was implemented to aggregate all observed rotations into

representative rotations. In all, nearly 170 unique crop rotations representing the entire U.S were

generated. The product is highly accurate with >95% accuracy for corn and soybean growing regions

in U.S. when compared to census data from NASS (USDA National Agricultural Statistical Services).

Crop rotations were then combined with data on soils, elevation, and hydrography to provide

historical input data for the biophysical and biogeochemical modeling. For future biofuel-scenario

analyses, current CDL land use classes are summarized into several major categories, including field

crops (FC), herbaceous vegetation (HV), and woody vegetation (WV), to reduce the number of

modeling units and the computational burden of SEIMF implementation.

5

2.3 Preparing and Implementing Spatially Explicit Simulations with the EPIC Model

2.3.1 The EPIC Model

The EPIC model can simulate the growth and development of over 100 plant species including all

major crops, grasses, legumes, and some trees (Williams, 1995). Crops can be grown as sole crops or

as intercrops (up to 10 species), in complex rotations, and under a wide range of management

operations including tillage, irrigation, fertilization, and liming. EPIC uses the concept of radiation-

use efficiency (Monteith, 1977) by which a fraction of daily photosynthetically active radiation is

intercepted by the plant canopy and converted into plant biomass. Daily gains in plant biomass are

affected by vapor pressure deficits and atmospheric CO2 concentration (Stockle et al., 1992 a,b). Plant

phenology is controlled via heat-unit calculations where each crop / plant species has base and

optimal air temperatures for growth. Potential daily gains in biomass are affected by environmental

stresses such as water, temperature, nutrients (primarily N and P), and aeration. All stresses are

calculated every day during the simulation, but only the value of the most severe stress is used to

reduce potential plant growth and crop yield. Stress factors for soil strength, temperature, and

aluminum toxicity are also calculated daily and used to adjust potential root growth (Jones et al.,

1991). EPIC is driven by daily weather consisting of solar radiation, air temperature, precipitation,

wind speed, and relative humidity, which is either predicted from weather parameters or read from

input files. Simulated processes are modified by topographical (e.g. slope gradient and length)

characteristics and field / watershed dimensions, soil layer properties (e.g., layer depth, bulk density,

C and N contents, pH), and management information (e.g. cropping systems, planting, fertilization,

irrigation, harvesting), as inputs to EPIC (Gassman et al., 2005; Williams et al., 1995; Zhang et al.,

2010).

For this work, the parameterization of the EPIC model was revised to ensure adequate

representation of biofuel cropping systems (Izaurralde et al., 2012). Specific crop parameters /

variables that underwent evaluation and eventual change included radiation use efficiency, root-to-

shoot ratio (Zhang et al., 2011), and planting density. Further, the rate of transformation of standing

live to standing dead vegetation was reduced from 1% to 0.1% per day to better model the reduced

loss of yield in fall-harvested cellulosic crops.

2.3.2 Spatially Explicit Modeling Framework

The high spatial resolution of operational modeling units (from tens to hundreds of meters,

with finer resolutions preferred) provides more homogeneous land units and improves model

accuracy. However, agricultural statistics (e.g. crop yields, nutrient applications, tillage types) are

only available at regional scale (county or state level). Therefore, a hierarchical data organization

system (HDOS) (Figure 2) was designed to facilitate fusing geospatial data from multiple sources

at different spatial resolutions, including climate input, land use, soil, and topography, to derive

homogeneous spatial modeling units (HSMU) at fine scale, while simultaneously allowing geo-

referencing of crop management information from different sources to specific HSMUs (Figure

3). The HDOS is designed to be flexible to use data at different levels of detail. Using the

conterminous U.S. as an example, the HDOS is illustrated in Figure 2. Given this HDOS, the

modeling results can be flexibly aggregated into several levels: county, state, and conterminous

6

U.S. modeling units. The finest resolution is

~ 60 m, which is determined by the

maximum resolution of the CDL and

SSURGO data. The HSMU is determined

by the unique combination of state, county,

LCLU, soil, and hydrologic catalogue unit

(Figure 3). Each HSMU has a unique ID

and the associated attribute variables that

allow preparing climate, land surface, and

management parameters for input to the

EPIC model.

For each HSMU, EPIC simulates all of

the land management scenarios. From the

output of these scenarios, a set of variables

(Table 2) represents the productivity and

environmental impacts of all possible

biofuel crop production systems. These

results are stored in 3 alternative formats to

facilitate multi-platform data access:

Microsoft Excel worksheets, Access databases, and PostgreSQL databases. Internet based data

transfer is also available by using PostgreSQL. These modeling results are linked to the HSMU

map by using the unique ID field (Figure 3), which allows users to visualize the spatial

distribution of the output variables.

2.3.3 Site Validation

Site-scale simulations based on specific data were conducted to calibrate the EPIC model

before applying it to the regional-scale simulations. Data from two Long-term Ecological

Research (LTER) sites at the Kellogg Biological Station (KBS) in Michigan and at Cedar Creek

(CDR) in Minnesota were used to parameterize EPIC and model aboveground net primary

productivity (ANPP) of successional vegetation growing on former agricultural fields. Data

included in the parameterization and initialization included historical weather, terrain

characteristics, and soil properties. The treatment selected for the simulation was natural

herbaceous vegetation or early successional (ES). The treatment was initiated in 1989 when

agricultural management ceased and natural vegetation was allowed to re-establish. Spring

burning was implemented in 1997 and whenever necessary thereafter to inhibit colonization of

woody species. Dominant species of successional vegetation included Solidago canadensis,

Elytrigia repens, Poa pratensis, Aster pilosus, Phleum pratense, Trifolium pratense, Apocynum

cannabinum, Daucus carota, Hieracium spp., and Rhus typhina (http://lter.kbs.msu.edu/).

Figure 2. Illustration of the hierarchical and flexible

downscaling and upscaling schemes in United States lands

(Zhang et al., 2010).

7

Figure 3. Schematic diagram describing the procedures for defining HSMUs (Homogenous

Spatial Modeling Units).

The EPIC calibration with KBS LTER results was achieved by simulating the composition of

the ES plant community with three species from the EPIC crop database (Poa pratensis, Phleum

pratense, and Trifolium pratense). Spring burning with 80% efficiency was simulated in 1997,

2003, 2004, 2006, and 2008. Aboveground NPP was simulated under three different scenarios;

one was left completely unmanaged (no harvest and no N addition), and the other two were

simulated with fall harvest with N at 0 and 123 kg ha-1

yr-1

. EPIC model runs were based on

historical records of daily maximum and minimum air temperature and precipitation from the

KBS LTER dataset (http://lter.kbs.msu.edu/datatables). Daily solar radiation, relative humidity,

wind speed, and missing temperature and precipitation were acquired from Gull Lake NWS

weather station (42º 24’ N 85º 23’ W) also at KBS. Soil layer properties used in the simulations

were those of the Kalamazoo soil series (fine-loamy, mixed, mesic Typic Hapludalf).

The EPIC simulations of ANPP for Cedar Creek, MN were conducted with data from the

CDR LTER site. A 60-yr chronosequence experiment was simulated according to Zak et al.

(1990). The chronosequence consisted of a series of 14 agricultural fields on Alfic and Typic

Udipsamments (Soil Survey Staff, accessed March 2012) left unmanaged during periods ranging

from one to 60 years. Floristic surveys conducted in 1987 revealed the presence of Ambrosia

artemisiifolia, Andropogon gerardii, Agropyron repens, Agrostris scabra, Polygonum

8

convolvulus, Poa pratensis, and Schizachyrium scoparium, among others. C3 species appeared to

dominate over C4 species during the first half of the chronosequence period while an equal mix

appeared to prevail during the second. The chronosequence was simulated with an equal mix of

generic C3 and C4 species from the EPIC crop database and driven by simulated daily weather

generated from data of weather station at Rosemount, MN near CDR LTER station. Weather

parameters were used to simulate ANPP during a 60-yr period intended to approximate

environmental conditions for 1928 – 1987. The only simulated external N addition was 7.7 kg

ha-1

yr-1

via wet deposition.

2.4 Regional Simulations

2.4.1 Simulation of Biomass Feedstock across the 10-State US Midwest

The EPIC-based Spatially Explicit integrative modeling framework (SEIMF) (Zhang et al.,

2010) was used to simulate yields of perennial herbaceous species grown on marginal lands

across the 10-state North Central USA study region. This region extends from North Dakota in

the NW corner, south to Nebraska in the SW corner, east to Ohio in the SE corner, and up to

Michigan in the NE corner. The Canadian – U.S. border defines the northern border of the study

region. A geospatial database containing soil, terrain, weather, land use/land cover and

management data was used to obtain relevant parameters for running the EPIC model.

Pertinent details of the geospatial database as described in Section 2.2 of this report follow:

a) Daily weather files at 32-km resolution to run EPIC were derived from the North America

Regional Reanalysis (NARR) (http://www.esrl.noaa.gov/psd/data/gridded/data.narr.html);

b) Elevation data (elevation, slope gradient and length) were extracted from the SRTM DEM

product while geospatial analysis was used to derive slope length and gradient of each HSMU.

c) Land use and land cover. Cropland data layer (CDL) for 2008 and SSURGO map were

combined to define LCLU and soil type at a spatial resolution of 60 m for the simulation domain.

Federal lands, golf courses, parks, large-lot single-family housing units, and vegetation planted in

developed settings for erosion control, recreation, or aesthetic purposes were excluded from the

simulation domain.

d) Soils and marginal lands. Soil layer properties to run EPIC originated from the SSURGO

database. Marginal lands were identified as rural lands falling into Land Capability Classes V-VII

with slope gradients <20% under non-forested vegetation.

Special consideration was given to the Sandhills of Nebraska (located in North Central

Nebraska) whose unique grass-stabilized sand dune topography distinguishes them from the

surrounding prairies (Eggemeyer et al., 2006). These dunes can be up to one hundred meters tall

and several kilometers long. The inter-dune valleys are the largest sources of hay for the cattle

industry in Nebraska (Gosselin et al., 2006). To keep only the inter-dune valleys and exclude the

fragile dune ridges and slopes from the analysis, we used the Topographic Position Index (TPI)

algorithm in ArcGIS (Tagil and Jenness, 2008).

9

2.4.2 Methodology for Siting Biorefineries

For each grid of marginal land, we simulated ANPP by the calibrated EPIC model under

three levels of N fertilization: 0, 68, and 123 kg N ha-1

yr-1

. The ANPP values obtained from

EPIC were used to identify the location of potential biorefineries that could process the cellulosic

feedstock. Only results belonging to an N fertilization rate of 68 kg N ha-1

yr-1

are used in this

analysis. The siting of a potential biorefinery was limited to areas where potential feedstock

derived from fields fertilized with 68 kg N ha-1

yr-1

could add at least 768 Gg yr-1

of cellulosic

biomass on marginal lands from within an economically feasible transportation distance of 80

km. To accomplish this, a moving window algorithm was implemented over the study region.

Subsequently, non-overlapping circles with the highest biomass yields were selected for potential

biorefinery locations. We used the following equation to convert biomass yields into liters of

ethanol:

Yield × 0.06 km × 0.06 km × 380.0 L Mg-1

× 100 = Yield × 136.8 (Eqn 1)

Where, Yield (Mg ha-1

) is biomass production, 0.06 × 0.06 km is the cell size of the model,

380.0 L Mg-1

is the conversion factor for converting cellulosic biomass to ethanol, and 100 is

number of hectares in 1 km2.

3.0 Results and Discussion

The results reported here address the two specific objectives described in Section 1.2: 1) a

spatially explicit national geodatabase to conduct biofuel simulation studies and 2) a regional analysis

of biorefinery siting based on cellulosic feedstock grown on marginal lands. They are also part of a

submitted manuscript examining the sustainability of bioenergy production from marginal lands in

the US Midwest (Gelfand et al., submitted)

3.1 Features of the National Geodatabase to Simulate Biofuel Production

The National Geodatabase of Contemporary Land Cover and Use in the Conterminous USA is

shown in Figure 4. The map is presented in highly aggregated form representing only some

predominant crop rotations (e.g. corn-based rotation, winter wheat – fallow) and land covers (e.g.

forests, grasslands, urban). This level of aggregation and abstraction is necessary in order to visualize

some of the predominant land uses and land covers. However, the map / database retains all the

richness

10

Figure 4. National Geodatabase of Land Cover and Use in the Conterminous USA.

Corn rotations

Urban and others

Grasslands

Forests

Winter Wheat/Fallow

11

of the high resolution data from which the map was built. In other words, it would be possible to

zoom in to a particular 60 m x 60 m field and see the details of its topography, soil properties, and

LCLU. Conversely, simulations conducted at the HSMU scale can be scaled-up to watershed, county,

state, and even national level without losing any information.

The level of detail included in this database would allow a user to exclude a number of features

not needed in the aggregation such as federal and provincial parks, protected areas, recreational areas,

urban, suburban, and commercial areas.

The crop rotation (land use) component is a unique feature of the database. Information on the

contemporary distribution of field crops on a given field and their change in sequence over time (i.e.

crop rotation) is crucial for the understanding of land issues such as soil quality, carbon management,

soil emissions of greenhouse gases, etc. As such, we believe the database presented here represents a

major advance for increasing the accuracy of environmental modeling and, in particular, for biofuel

modeling.

3.2 Simulations Results

3.2.1 Site Simulations

Aboveground NPP simulated EPIC calibration during 2000 and 2008 averaged 60.0±11.0 kg ha-1

yr-1

for the scenario without biomass harvest and fertilization, which is similar to the observed

average of 59.8±4.4 kg ha-1

(Table S6). Similarly, simulated ANPP for the scenario with harvest and

fertilization was very close to field-based estimation, 96±8 vs. 87.8±6.4 kg ha-1

yr-1

.

Direct comparisons of biomass productivity results are not possible for the CDR data because the

simulations are not time specific. Simulated standing aboveground biomass during the July-August

period averaged 51.3±11.3 kg ha-1

during the first three years. In comparison, Zak et al. (1990),

observed 50.9 kg ha-1

in the 1-yr old field and 29.6 kg ha-1

in the 3-yr old field. Simulated biomass

during the last three years of the simulation averaged 76.7±12.7 kg ha-1

, while Zak et al. (1990)

observed 78.3 kg ha-1

in the 60-yr old field. While EPIC failed to capture the decrease in plant

productivity observed during the first years of the chronosequence, it did capture the increase in

productivity toward the end of the simulation period. Similar to observations, simulated annual

biomass productivity correlated moderately well with N mineralization and soil N. However,

correlations obtained with observed data (R2 = 0.78 for potential N mineralization; R

2 = 0.94 for soil

N) (Zak et al., 1990) were considerably stronger than those obtained with simulated data (R2 = 0.34

for net N mineralization; R2 = 0.32 for soil N).

Overall, EPIC appeared to capture adequately the observed plant productivity patterns from both

KBS and CDR LTER sites under different environmental constraints and management scenarios. In

addition, Izaurralde et al. (2006) used EPIC for simulations of the effects of unmanaged grasses on

soil carbon accrual on marginal lands in Nebraska, Kansas, and Texas and found good agreement

between EPIC predictions and long-term field data, which provides further justification for using the

model for large-scale simulations.

12

3.2.2 Potential Biomass Production from Marginal Lands

The available acreage of marginal lands varied considerably by state (Figure 5). Overall,

available acreage of marginal lands varied between <100,000 ha in Ohio (1% of total land) to

more than 6.4 million ha in Nebraska (30% of total land). Nebraska was also the only state where

marginal land acreage exceeded that of arable lands.

State

IL IN IA MI MN NE ND OH SD WI

Ma

rgin

al L

an

ds (

ha

x 1

06)

0

1

2

3

4

5

6

7

Ara

ble

la

nd

s (

ha

x 1

06)

0

5

10

15

20

25

Marginal lands acreage

Arable lands acreage

Figure 5. Marginal lands acreage (×106 ha) in 10-state US North Central region

We simulated ANPP by the calibrated EPIC model under three levels of N fertilization: 0, 68, and

123 kg N ha-1

yr-1

, and two levels of harvest efficiency: 55% and 90%. Modeled field size for the

estimation of potential productivity varied substantially, with a minimum size of 3,600 m2 and a

maximum >650×106 m

2. Overall, we simulated 78,184 fields having unique combinations of soil

type, land-use, and LCC. The connectivity between each field and nearby fields was not assessed.

Average biomass yields (Mg ha-1

yr-1

) for 78,184 parcels of marginal land as modeled by EPIC for

three levels of N fertilizer application and two harvest efficiencies are presented in Table 1. Values in

parentheses are standard deviations. Ethanol yields are based on a conversion factor of 380.0 L Mg-1

of cellulosic biomass.

13

Table 1. Average biomass yields (Mg ha-1 yr-1) for 78,184 parcels of marginal land as modeled

by EPIC for three levels of N fertilizer application and two harvest efficiencies.

Average Yields

Biomass Ethanol

Mg ha-1

yr-1

L×103 ha

-1 yr

-1

EPIC model

Harvest efficiency (%)

55 4.3 (1.7) 1.6 (0.6)

90 6.0 (2.6) 2.3 (1.0)

N Fertilization (kg ha-1

yr-1

) †

0 6.0 (2.6) 2.3 (1.0)

68 8.2 (2.7) 3.1 (1.0)

123 8.9 (2.8) 3.3 (1.0)

† Simulations under different N fertilization regimes with assumed 90% harvest efficiency.

3.2.3 Comparison with Estimates from Billion Ton Study Update

We also compared our results of biomass production rates with recent Billion-Ton Study

Update (Perlack et al. (2011) (Table 2). Our estimation differs significantly and exhibited no

relation to the Billion-Ton Study Update (R2 = 0.004). The Spearman rank-order correlation

coefficient (rs) is -0.0061 with a p-value of 0.98. An rs value below zero implies negative

agreement between the two rankings.

Table 2. EPIC simulated and Billion-Ton Study Update estimated biomass yields on marginal

lands across the 10-state US Midwest Region.

Average biomass yield

EPIC* BTS-U

----------- Mg ha-1

-----------

Illinois 8.16 12.52

Indiana 8.35 13.07

Iowa 7.81 11.73

Michigan 9.98 7.44

Minnesota 10.63 6.04

Nebraska 6.01 8.87

North Dakota 5.55 4.59

Ohio 7.16 14.03

South Dakota 4.75 7.59

Wisconsin 8.96 10.29

* EPIC simulated for unfertilized marginal land.

The difference between our EPIC simulated yields and the results from the Billion-Ton

study update can be attributed to three different factors: First, by simulating the production of

14

Figure 5. Potential cellulosic ethanol biorefinery locations in the 10-state US Midwest Region.

mixed species assemblages of existing vegetation, we are able to avoid carbon debt associated

with the establishment of new perennial plantings on land not now farmed. Further, we

simulate production only on marginal lands rather than to a combination of all marginal and up

to 25% of current cropland. Third, rather than using a statistical model based on field data

scaled by growing season precipitation and temperature, we employ a process-level crop model

operating at a fine grained (0.4 ha) scale to fully capture the soil and climate interactions.

3.2.4 Identifying Location of Biorefineries in the 10-State US Midwest Region

We implemented a moving window algorithm to assess availability of cellulosic biofuel from

marginal lands in the 10-State US Midwest region. The placement of biorefineries for cellulosic

biomass production from marginal lands in ten North Central states is shown in Figure 6. Each circle

represents an 80 km radius area with sufficient biomass resources to produce at least 89.3 ML ethanol

yr-1 based on quantitative simulation of yields from non-forested marginal lands at a 60×60 m

resolution. Corn acreages (in ha) by county have been derived from the CDL of 2008. The inset map

shows a close up of a potential biorefinery location where the yield is not clearly visible on the main

map. Figure 5 shows the placement of biorefineries for cellulosic biomass production from marginal

lands in ten North Central states with identification numbers corresponding to Figure 6. Overall,

nearly 21 GL of cellulosic ethanol can be produced (Table 3), meeting ~30% of the 2022 target for

cellulosic biofuel mandated by the 2007 US Energy Independence and Security Act.

15

Table 3. EPIC modeled biomass and cellulosic ethanol production by state assuming 90% harvest

efficiency and two N-fertilization levels (0 and 68 kg ha-1

).

Total biomass production

Total cellulosic ethanol

production

ID§ State

† F*=0 F=68 F=0 F=68

-------- Mg × 106 -------- -------- GL yr

-1 --------

1 Illinois‡ 1.35 1.88 0.18 0.26

2 Illinois 1 0.67 0.8 0.09 0.11

3 Indiana‡ 1.18 1.43 0.16 0.19

4 Indiana 1‡ 0.77 0.96 0.11 0.13

5 Iowa 1.67 2.27 0.23 0.31

6 Iowa 1‡ 0.87 1.17 0.12 0.16

7 Michigan 1.78 2.29 0.24 0.31

8 Minnesota 2.38 2.55 0.32 0.35

9 Minnesota 1 0.86 1.10 0.12 0.15

10 Minnesota 2 0.90 1.11 0.12 0.15

11 Nebraska 13.72 19.13 1.88 2.62

12 Nebraska 1‡ 10.44 14.66 1.43 2.01

13 Nebraska 2‡ 9.75 13.83 1.33 1.89

14 Nebraska 3 8.89 11.54 1.22 1.58

15 Nebraska 4 6.43 9.01 0.88 1.23

16 Nebraska 5‡ 4.05 5.33 0.55 0.73

17 Nebraska 6 2.55 3.40 0.35 0.46

18 Nebraska 7‡ 1.68 2.30 0.23 0.31

19 Nebraska 8 0.61 0.80 0.08 0.11

20 North Dakota 5.48 7.50 0.75 1.03

21 North Dakota 1 4.01 5.19 0.55 0.71

22 North Dakota 2 3.75 5.12 0.51 0.7

23 North Dakota 3 3.59 4.62 0.49 0.63

24 North Dakota 4‡ 1.52 1.89 0.21 0.26

25 North Dakota 5 1.28 1.80 0.17 0.25

26 North Dakota 6‡ 0.74 0.95 0.10 0.13

27 South Dakota 7.22 10.97 0.99 1.50

28 South Dakota 1‡ 4.25 6.07 0.58 0.83

29 South Dakota 2 1.97 2.77 0.27 0.38

30 South Dakota 3‡ 1.86 2.42 0.26 0.33

31 South Dakota 4‡ 0.93 1.22 0.13 0.17

32 Wisconsin 1.64 2.03 0.22 0.28

33

34

Wisconsin 1‡

Wisconsin 2‡

Total

1.20

0.53

110.52

1.58

0.65

150.34

0.16

0.07

15.1

0.22

0.09

20.57 * N fertilizer application in kg N ha

-1 yr

-1.

16

† If more than one potential biorefinery could be placed in a state, we name potential placements with serial

numbers. ‡ Biomass collection for the biorefinery location involves collection across state boundaries.

§ Biorefinery placements associated with biorefinery ID are presented in Figure 6.

Figure 6. The placement of biorefineries for cellulosic biomass production from marginal lands in

ten North Central states with identification numbers corresponding to Table 3.

3.3 Perennial Biomass Crop Production in the Sandhills of Nebraska

The dune ridges and slopes of the Sandhills of Nebraska were removed from analysis by applying

the TPI algorithm available in ArcGIS. The TPI is a classification scheme based on the difference in

elevation values between a cell in a DEM raster and its neighbors. The extent to which a cell is higher

or lower as compared to its neighbors, combined with its slope, can be used to assign it a landform

classification like valley, ridge etc.

Removing the dune ridges and slopes reduced the area available for consideration as marginal

lands in Nebraska by more than 200,000 ha. This constitutes nearly 3% of the marginal lands

identified by our approach. The inter-dune valleys falling under LCC V-VII and with slope gradients

< 20% were then used for further analysis using EPIC model with yields shown in Figure 7.

17

Figure 7. Production capability across the Sandhills for a native prairie mix. The dune ridges and

slopes of the Sandhills have been excluded from analysis.

18

4.0 Conclusions

We have developed a spatially explicit geodatabase containing data for 10 of the US Midwest

states for the primary purpose of conducting simulation studies of potential biofuel scenarios. In

addition, we have used this geodatabase to perform simulation studies in these states using the

EPIC biogeochemical model. The results of this study will be submitted to the USDOE

Bioenergy Knowledge Discovery Framework as a way to contribute to the development of a

sustainable bioenergy industry.

In order to perform these high-resolution analyses, it was vital to have a computing

infrastructure in place that would allow us to simulate 78,184 individual land units for each of the

three fertilization and two harvest efficiency scenarios and would allow us to aggregate and

visualize the resulting data. The first part of this infrastructure was the server “deltac”: an eight-

core Linux server, acquired with ARRA support, used to conduct the initial county-level and

state-level simulations and host the PostgreSQL database where data from the national-scale

simulations were stored. Had we not received funding for this vital resource, this report would not

have been possible in the time allotted. ARRA funding also supported a PhD student from UMD

who worked on creating the geodatabases and executing the simulations in this study.

While the available acreage of marginal lands varied considerably by state, we determined

that 34 potential biorefineries could be situated among 10 of the US Midwest states, each of

which could produce at least 23.6 million gallons ethanol yr-1

(Error! Reference source not

ound.). EPIC simulated annual yields for all marginal lands in the region averaged 6.0±2.6 Mg

ha-1 (Table 9). Modest fertilization rates of 68 kg N ha-1

yr-1

increased estimated average yields

by 36% to 8.2±2.7 Mg ha-1. Spatially explicit biophysical modeling of cellulosic feedstock

production on marginal lands within 80 km of a potential biorefinery suggests an annual potential

ethanol production of ~5.6 billion gallons yr-1

, or ~30% of the 2022 target for cellulosic biofuel

mandated by US legislation.

In summary, we have reported on the development of a spatially explicit national geodatabase

to conduct biofuel simulation studies and provided simulation results on the potential of perennial

cropping systems to serve as feedstocks for the production of cellulosic ethanol. Our results show

that management of marginal lands to permit the growth of mixed species assemblages can

provide cellulosic biofuel feedstock without creating the carbon debt associated with land use

change to new plantings. This would also lead to additional GHG benefits because of lesser

dependence on N inputs and higher amounts of soil C sequestration. This work, we believe, opens

the door for further analysis on the characteristics of cellulosic feedstocks as major contributors to

the development of a sustainable bioenergy economy. Examples of research questions that could

be pursued with the modeling framework presented here include:

How can the modeling framework be improved? (e.g. adding irrigation, improving

winter wheat simulations)

What is the performance of emerging biofuel feedstocks such as miscanthus, energy

cane, and energy sorghum? Where are the best regions to grow them?

19

What is the potential of marginal lands across the conterminous USA to provide

sustainable levels of biomass feedstocks to the cellulosic ethanol industry?

What are the full GHG impacts of diverse biofuel production systems?

20

5.0 References

Crutzen P. J.; A. R. Mosier; K. A. Smith, and W. Winiwarter. 2008. N2O release from agro-

biofuel production negates global warming reduction by replacing fossil fuels. Atmos. Chem.

Phys. 8:389-395.

Dale, B.E., and S. Kim. 2004. Global potential bioethanol production from wasted crops and crop

residues. Biomass Bioenergy 26:361-375.

Egbendewe-Mondzozo A., S.M. Swinton, R.C. Izaurralde, D.H. Manowitz, and X. Zhang.

Biomass supply from alternative cellulosic crops and crop residues: A spatially-explicit

bioeconomic modeling approach. Biomass Bioenergy doi:10.1016/j.biombioe.2011.09.010.

Eggemeyer, K.D., T. Awada, D.A. Wedin, F.E. Harvey, and X. Zhou. 2006. Ecophysiology of

two native invasive woody species and two dominant warm-season grasses in the semiarid

grasslands of the Nebraska Sandhills. http://digitalcommons.unl.edu/natrespapers/38/.

EISA [Energy Independence and Security Act]. 2007. H.R. 6 110th Congress of the United States

of America, Government Printing Office, Washington, DC. 309 pp.

Fargione J., J. Hill, D. Tilman, S. Polasky, and P. Hawthorne. 2008. Land clearing and the biofuel

carbon debt. Science 319:1235–1238.

Farr, T.G., P.A. Rosen, E. Caro, et al. 2007. The shuttle radar topography mission. Review of

Geophysics 45, RG2004 doi:10.1029/2005RG000183.

Gassman, P.W., J. R. Williams, V.W. Benson, R.C. Izaurralde, L.M. Hauck, C.A. Jones, J.D.

Atwood, J.R. Kiniry, and J.D. Flowers. 2005. Historical development and applications of the

EPIC and APEX models. CARD Working Paper 05-WP. Center for Agricultural and Rural

Development, Iowa State University, Ames, Iowa, USA. Available at:

http://www.card.iastate.edu/publications/synopsis.aspx?id=763.

Gosselin, D. C., V. Sridhar, F. E. Harvey, and J. W. Goeke. 2006. Hydrological effects and

groundwater fluctuations in interdunal environments in the Nebraska Sandhills. Great Plains

Research 16: 17-28.

Izaurralde, R.C., J.R. Williams, W.B. McGill, N.J. Rosenberg, and M.C. Quiroga Jakas. 2006.

Simulating soil C dynamics with EPIC: Model description and testing against long-term data.

Ecol. Modelling 192:362-384.

Izaurralde, R.C., W.B. McGill, and J.R. Williams. 2012. Development and application of the

EPIC model for carbon cycle, greenhouse-gas mitigation, and biofuel studies. p. 409-429 In

A. Franzluebbers, R. Follett, and M. Liebig (eds.) Managing Agricultural Greenhouse Gases:

Coordinated agricultural research through GRACEnet to address our changing climate,

Elsevier, Amsterdam.

Jones, C.A., P.T. Dyke, J.R. Williams, J.R. Kiniry, V.W. Benson, and R.H. Griggs. 1991. EPIC:

an operational model for evaluation of agricultural sustainability. Agric. Syst. 37:341-350.

Kiniry, J.R., D.J. Major, R.C. Izaurralde, J.R. Williams, P.W. Gassman, M. Morrison, R.

Bergentine, and R.P. Zentner. 1995. EPIC model parameters for cereal, oilseed, and forage

crops in the northern Great Plains region. Can. J. Plant Sci. 75:679-688.

Klingebiel, A.A., and P.H. Montgomery. 1961. Land Capability Classification. Agriculture

Handbook No. 210, Soil Conservation Service, U.S, pp. 1–3. Department of Agriculture,

Washington, DC.

Monteith, J.L., 1977. Climate and the efficiency of crop production in Britain. Philos. Trans. R.

Soc. London B 281:277-294.

Melillo, J.M., J.M. Reilly, D.W. Kicklighter, A.C. Gurgel, T.W. Cronin, S. Paltsev, B.S. Felzer,

X.D. Wang, A.P. Sokolov, and CA Schlosser. 2009. Indirect Emissions from Biofuels: How

Important? Science 326:1397-1399.

21

NRC, 2011. Renewable fuel standard: Potential Economic and Environmental effects of U.S.

Biofuel Policy. Committee on Economic and Environmental Impacts of Increasing Biofuels

Production, National Research Council, National Academy Press, Washington, DC. 423 pp.

Robertson, G.P., V.H. Dale, O.C. Doering, S.P. Hamburg, J.M. Melillo, M.M. Wander, W.J.

Parton, P.R. Adler, J.N. Barney, and R.M. Cruse, C.S. Duke, P.M. Fearnside, R.F. Follett,

H.K. Gibbs, J. Goldemberg, D.J. Mladenoff, D. Ojima, M.W. Palmer, A. Sharpley, L.

Wallace, K.C. Weathers, J.A. Wiens, and W.W. Wilhelm. 2008. Sustainable biofuels redux.

Science 322:49-50.

Searchinger, T., R. Heimlich, R.A. Houghton,F. Dong, A. Elobeid, J. Fabiosa, S. Tokgoz, D.

Hayes, and T.H. Yu.2008. Use of U.S. Croplands for biofuels increases greenhouse gases

through emissions from land-use change. Science 319:1238-1240.

Simley, J.D., and J.W.J. Carswell. 2009. The National Map – Hydrography. U.S. Geological

Survey Fact Sheet 2009-3054, pp 4. Available at: http://pubs.usgs.gov/fs/2009/3054/

(accessed February 2011).

Slater, S., K. Keegstra, and T.J. Donohue. 2010. The US Department of Energy Great Lakes

Bioenergy Research Center: Midwestern biomass as a resource for renewable fuels.

Bioenergy Res. 3:3-5.

Soil Survey Staff, Natural Resources Conservation Service, United States Department of

Agriculture (USDA). Available at: http://soildatamart.nrcs.usda.gov. Accessed March 20,

2012.

Stockle, C.O., J.R. Williams, C.A. Jones, and N.J. Rosenberg. 1992a. A method for estimating

the direct and climatic effects of rising atmospheric carbon dioxide on growth and yield of

crops. I. Modification of the EPIC model for climate change analysis. Agric. Syst. 38:225-

238.

Stockle, C.O., P.T. Dyke, J.R. Williams, C.A. Jones, and N.J. Rosenberg. 1992b. A method for

estimating the direct and climatic effects of rising atmospheric carbon dioxide on growth and

yield of crops. II. Sensitivity analysis at three sites in the midwestern USA. Agric. Syst.

38:239-256.

Tagil, S., and J. Jenness. 2008. GIS-based automated landform classification and topographic,

landcover and geologic attributes of landforms around the Yazoren Polje, Turkey. Journal of

Applied Sciences 8: 910–921.

U.S. Department of Energy. 2011. U.S. Billion-Ton Update: Biomass Supply for a Bioenergy and

Bioproducts Industry. R.D. Perlack and B.J. Stokes (Leads), ORNL/TM-2011/224. Oak

Ridge National Laboratory, Oak Ridge, TN. 227pp.

Williams, J.R., C.A. Jones, J.R. Kiniry, and D.A. Spanel. 1989. The EPIC crop growth model.

Trans. ASAE 32:497-511.

Williams, J.R. 1995. The EPIC model. In Computer Models of Watershed Hydrology (Ed.: V.P.

Singh). Water Resources Publications, Highlands Ranch, CO. pp 909-1000.

Zak, D.R., D.F. Grigal, S. Gleeson, and D. Tilman. 1990. Carbon and nitrogen cycling during old

field succession: Constraints on plant and microbial biomass. Biogeochemistry 11:111-129.

Zhang, X., R.C. Izaurralde, D. Manowitz, T.O. West, W.M. Post, A.M. Thomson, V.P. Bandaru,

J. Nichols, and J.R. Williams. 2010. An integrative modeling framework to evaluate the

productivity and sustainability of biofuel crop production systems. Global Change Biol. –

Bioenergy 2:258–277.

22

6.0 Publications and Presentations

Bandaru, V., W. Post, S. Kang, T. West, R. Izaurralde, and X. Zhang. Integrated Assessment

Approach to Identify Potential Marginal Lands for Alternative 2010. Bio-Fuel Crop

Production. ASA, CSSA, and SSSA 2010 International Annual Meetings, Oct. 31 – Nov. 4,

Long Beach, CA.

Egbendewe-Mondzozo, A, SM Swinton, C Izaurralde, D Manowitz, and X Zhang. 2011. Biomass

supply from alternative cellulosic crops, and crop residues: A watershed scale bioeconomic

modeling approach. Ecological Society of America, Annual Meeting, Austin, TX.

Egbendewe-Mondzozo, A., S.M. Swinton, R.C. Izaurralde, D.H. Manowitz and X. Zhang. 2011.

Biomass Supply From Cellulosic Crops and Crop Residues in Southwest Michigan. GLBRC

2011 Retreat, Kalamazoo, MI.

Gelfand, I., R. Sahajpal, X. Zhang, R.C. Izaurralde, and G.P. Robertson. 2010. Biofuel production

and climate mitigation potential from marginal lands in US North Central region. 2010 AGU

Fall Meeting, San Francisco, CA.

Gelfand, I., R. Sahajpal, X. Zhang, R.C. Izaurralde, K.L. Gross, and G.P. Robertson. Sustainable

bioenergy production from marginal lands in the US Midwest. (Submitted).

Izaurralde, R.C., X. Zhang, D.H. Manowitz, and R. Sahajpal, Water-cycle implications of biofuel

production systems, GLBRC 2011 Retreat, May, 2011, South Bend, IN.

Izaurralde, R.C., X. Zhang, Sustainable biofuel operating space: multi-criteria assessment and

multi-objective optimization. GLBRC 2011 Retreat, May, 2011, South Bend, IN.

Izaurralde, R.C., W.B. McGill, and J.R. Williams. 2012. Development and application of the

EPIC model for carbon cycle, greenhouse-gas mitigation, and biofuel studies. p. 409-429 In

A. Franzluebbers, R. Follett, and M. Liebig (eds.) Managing Agricultural Greenhouse Gases:

Coordinated agricultural research through GRACEnet to address our changing climate,

Elsevier, Amsterdam.

Kang, S, WM Post III, J Nichols, D Wang, C Izaurralde, TO West, P Bandaru, X Zhang, D

Manowitz, and AM Thomson. 2011. Delineating a hierarchal definition of marginal land

using simulation results. Ecological Society of America, Annual Meeting, Austin, TX.

Kang, S., J. Nichols, D. Wang, W.M. Post, R.C. Izaurralde, X. Zhang, D.H. Manowitz, and V.

Bandaru. Using high-performance biophysical simulations for sustainability assessment and

design of bioenergy production systems: a preliminary case study in RIMA of Wisconsin.

GLBRC 2011 Retreat, May 2011, South Bend, IN.

Kang, S., V. Bandaru, W. Post, T. West, J. Nichols, R. Izaurralde, X. Zhang, A. Thomson, and D.

Manowitz. 2010. Marginal Land Assessment and Management for Bioenergy Production: A

Case Study at RIMA of Michigan. ASA, CSSA, and SSSA 2010 International Annual

Meetings, Oct. 31 – Nov. 4, Long Beach, CA.

Manowitz, D and RC Izaurralde. 2011 Modeling production, net greenhouse gas emissions, and

related environmental impacts of bioenergy systems at plot scale. Ecological Society of

America, Annual Meeting, Austin, TX.

Nichols, J.A, S. Kang, D. Wang, W.M. Post, R.C. Izaurralde, X. Zhang, D.H. Manowitz, and V.

Bandaru. HPC-EPIC for high resolution simulation of environmental and sustainability

assessment, GLBRC 2011 Retreat, May 2011, South Bend, IN.

Sahajpal, R, TD Meehan, RC Izaurralde, D Manowitz, X Zhang, C Gratton, BP Werling and DA

Landis. 2011. Integrating biodiversity estimation into the EPIC model. Ecological Society of

America, Annual Meeting, Austin, TX.

Sahajpal, R., R.C. Izaurralde, and X. Zhang. 2010. Siting cellulosic ethanol biorefineries based on

perennial biomass crop production on marginal lands in the 10-State U.S. North Central

Region. GLBRC 2010 Retreat, South Bend, IN.

23

Sahajpal, R., R.C. Izaurralde, and X. Zhang. Identifying Marginal Lands Suitable for Cultivation

for Perennial Biomass Crop Production in the Sandhills of Nebraska, GLBRC 2011 Retreat,

May, 2011, South Bend, IN.

Thomson, A.M, R.C. Izaurralde, K.V. Calvin, X. Zhang, M.A. Wise, and T.O. West. 2010.

Integrating a Detailed Agricultural Model in a Global Economic Framework: New methods

for assessment of climate mitigation and adaptation opportunities. 2010 AGU fall meeting,

San Francisco, CA.

Wang, D., Nichols, J., S. Kang, W.M. Post, R.C. Izaurralde, X. Zhang, D.H. Manowitz, and V.

Bandaru. Large-scale data management system for spatially-explicit agroecosystem

modeling. GLBRC 2011 Retreat, May 2011, South Bend, IN.

Zhang, X., R.C. Izaurralde, J.G. Arnold, N.B. Sammons, D.H. Manowitz, A.M. Thomson, J.R.

Williams, (2011) Comment on “Modeling Miscanthus in the Soil and Water Assessment Tool

(SWAT) to Simulate Its Water Quality Effects As a Bioenergy Crop”. Environ. Sci. Tech.

45:6211-6212.

Zhang, X, RC Izaurralde, JG Arnold, NB Simmons, D Manowitz, AM Thomson, and J Williams.

2011. SEIMF: A spatially-explicit integrative modeling framework to evaluate the

productivity and sustainability of biofuel crop production systems. Ecological Society of

America, Annual Meeting, Austin, TX.

Zhang, X., R.C. Izaurralde, A.M. Thomson, R. Sahajpal, D.H. Manowitz, and S.D. LeDuc,

Assessment of Conservation Reserve Program (CRP) and Marginal Land-Use Change for

Bioenergy Production. 2011 AAG Annual Meeting, April, 2011, Seattle, WA.

24